Michelle: FES work focused on operational concerns.Tested critical deployment tasks like tenant provisioning, major release upgrades.
Tested migration and deployment of Eureka clusters.
Tested configuration options of Keycloak. Roles are to give the user a system role. FOLIO modules remain at the core of the platform. Eldliiar: Eureka Development in Rancher Historically, Okapi was single point of failure. On Eureka, things will change. Keycloak and Kong can serve multiple clusters, as opposed to Okapi. Florian: Which parts of the UI are visible ? How does that intersect with Stripes ? Eldiiar: There are small adjustments on top of Stripes file. Once we have built UI, we do not need to rebuild it anymore. All (configurations) are permanent. And configurable via the Keycloak API. There is a graphical user interface. Florian K.: We have a Reshare model. How much work will it be to adjust to the new platform ? Eldiiar: It is just one commit and several lines. You will gain another opportunity. Everything will work like a charme. Craig: Nothing to add. Eldiiar: If Okapi is down, the system is not accessible. In Eureka, if Kong is down, backend modules are still able to communicate via Sidecars. Florian K. : But you have the problem when Keycloak goes down. Eldiiar: So far we have never had problems with Keycloak and Kong. Clusterisation of Okapi was hard to achvieve. You had to deal with Hazelcast. With Kong and sidecars it is already clustered. Everything works like a charm. - modern and fast app gateway (Kong) - Envionment provisioning duration decreases from ~45 to ~30 minutes. - sidecars supports default log level Mark: "It (Kong) takes two cups of coffee per hour instead of one" (laughter) Jason: How many URLs do you provide ? Eldiiar: We just provide a single Kong URL for the whole process. Kong is just an application gateway. Everything happens under the hood (of Sidecars). Jason: In Hazelcast, ... Craig: The modules are unaware of ...; Everything is handled by the sidecars. Mark: Kafka used Hazelcast; we have replaced Zookeeper by Graft. Denis Mönch: ... (question about logging) Answer: Sidecar logging is disctinct from ... Uwe Reh: Are sidecars and modules combined in a pod ? Edliiar: Yes. So, it doesn't double the number of pods. The demo links (on the last slide) run on the same Kubernetes cluster on the same namespace. They are being destroyed on daily basis. Tobias: Direct module to module communication. Was that written inhouse ? I do not understand the purpose of Quarkus. Craig: Module to module communication has been provided by the platform. It is part of the business logic. Tobias: What makes available the specific module code ? Is that Quarkus or what is the purpose of Quarkus ? Craig: The module needs to call another module. It takes its requests to the sidecar. The sidecar routes the request. Eldiiar: The sidecar pulls down all communication out of discovery and keeps it in memory. So, even if Kong is not available, the information is there. The sidecar does the routing and communication. Owen: What about timeouts ? This is handled at the Okapi level at the moment. Craig: Sidecars are built on Quarkus, a java framework. Eldiiar: With the sidecars, we can configure our environment more granuarly. Next link is to our public available Helm charts. folio-org/folio-helm-v2 Pipeline example: https://shorturl.at/aNXyq Uwe/ Josh: single server installations ... Mark: Thank you Hong Wei et. al.
End of presentation 16:42 o'clock
Questions, Discussion
Jason: I am still digesting.
Ingolf: Is there a value in going through the eureka-platform-bootstrap example if one plans to use Kubernetes for deployment ? Likewise for Okapi deployment, years ago, we first went through the single server example at then, with that experience, started deploying on Kubernetes. Do you recommend going through this example ? It is based on docker-compose.
Craig : https://github.com/folio-org/eureka-platform-bootstrap Craig: The platform-bootstrap example was conceived in order to save AWS costs for the development environments. Going through the eureka-platform example (as a system administrator) is not a requirement. Ian Walls: is there a level minimum ...
Uwe: Is there any documentation about common pitfalls ? Josh: Keycloak in a single namespace ? A seperate Keylcoak in another namespace ?
Eldiinar: You can utilise your existing Keycloak.
Tobias: A private Repository for AWS. Where to find docuemntation about building the manager applications, for example. Will you provide it on folio-org ? Answer: This is the first time that I heard about problems with building. Eldiiar: Yes, we will have the images on Dockerhub. They will be prebuilt.